Fixed shower shapes

[bendavid]

Combines and supercedes
#1510
#1502

Makes all cluster shape computations inherit rechit cleaning from underlying cluster and consistently use rechit fractions, both for eNxM style quantities as well as covariance and localCovariance (sigIetaIeta etc).

[cmsbuild]

+1
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/1400/summary.html

[cmsbuild]

A new Pull Request was created by @bendavid (Josh Bendavid) for CMSSW_7_0_X.

Fixed shower shapes

It involves the following packages:

RecoEcal/EgammaCoreTools
RecoEgamma/EgammaPhotonProducers
RecoEgamma/EgammaElectronAlgos

@thspeer, @slava77 can you please review it and eventually sign? Thanks.
You can sign-off by replying to this message having '+1' in the first line of your reply.
You can reject by replying to this message having '-1' in the first line of your reply.
@ktf you are the release manager for this.

[cmsbuild]

Pull request #1519 was updated. @nclopezo, @cmsbuild, @thspeer, @slava77 can you please check and sign again.

[bendavid]

n.b. The only change here was to rebase on top of CMSSW_7_0_0_pre8 in order to be consistent with Egamma GED development recipes.

[cmsbuild]

+1
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/1433/summary.html

[slava77]

This is a first piece.
A summary of changes in the regular (non-GED) mode.
@bendavid @lgray
Josh, Lindsey, please check if this is expected

Tested (as sign274) in CMSSW_7_0_X_2013-11-22-0200 in combination
Fixed shower shapes
- #1519 f3395c4

• add configurable cut on clusters to be added to the event
  • add configurable cut on clusters to be added to the event #1576 9c44139

All changes should be from #1519

Regular mode:
differences are observed pretty much everywhere related to egamma

• minor changes in momentum assignment
• quite significant changes in cluster shape variables (as expected)
  • e1x5 and e5x5 appear to be more broad
  • sigmaEtaEta is more narrow

Details:

• 16.0 ele 1TeV
  • small diffs in energies, mostly at lower values (e1x5, e5x5)
• 17.0 ele 35
  • momentum is pretty much unchanged for electrons
  • significant changes in e1x5 and similar quantities
    
    
    
  • regular photons et are affected less (fractions are not used here, the minor diffs are probably from dropping the flags. is this expected?)
    
    
• 18.0 gamma 10 GeV
  • improvement from using fractions is most notable for gedPhotons (it looks really more like a bug fix here than a feature). The et is centered at 10GeV pretty nicely
    
    
    
• 19.0 gamma 35 GeV
  • the effect on gedPhotons is much weaker here. The mean of rec/true actually drifts down away from 1
    
    
• 202.0 TTBar PU (mostly fake eles/gammas)
  • significant change (smaller now) in the sigmaEtaEta for electrons, similar to above
    

[bendavid]

Hi Slava,
The changes in shower shapes look reasonable and more or less as expected. Do you have the plots also of R9 (for both standard and GED). One of the main technical points here is that R9 should be strictly <=1.0 now.

For the small change in momentum assignment for non-ged photons, this arises because of a hard switch in the old photon reco where energy assignment depends on E5x5 directly for photons with R9>0.94(0.95) in barrel (endcap). So the small changes in e5x5 and e3x3 are expected to have some small effect here.

For the change in momentum assignment for the low Et GED photons I will discuss briefly with @lgray.

[slava77]

I can't say much about R9

DQM plot is not binned finely enough

.. and I can't make a plot of this in fwlite, because the variable is derived from a value pointed by a Ref (->superCluster()->rawEnergy()), which doesn't work with simple TTree::Draw

[slava77]

part two of the summary, now in the GED mode:

• electron gun (wflow 16 and 17) are similar to the non-GED
• 18.0 gamma 10
  • particleFlowEgamma distribution shows a pretty large drop of candidates
    Compare the change in regular (non-GED) mode [only the first plot is not in GED mode]
    
    with the GED mode
    
    
  • the downstream pf candidates look fairly unchanged (is it consistent?)
  • the gedPhotons have a similar drop in yield (note that the regular mode also has a drop in yields, not as significant here)
    
    
• 19.0 has no drop as above. Changes in shapes consistent with the regular mode

[bendavid]

Hi Slava,
The change in momentum assignment for the GED Photons is understood. They still inherit the old R9-dependent energy assignment for now. (But this will imminently change when the regression is made the default)

The change in the number of candidates for the gamma10 sample is expected given that the Et cut on the photons in the reconstruction is 10GeV, so one sits right on the edge, and small changes to energy response will change the efficiency. (Meaning one should rather use a harder photon gun sample to validate in fact.)

[slava77]

+1

Tested (as sign274) in CMSSW_7_0_X_2013-11-22-0200 in combination

• Fixed shower shapes
  • Fixed shower shapes #1519 f3395c4
• add configurable cut on clusters to be added to the event
  • add configurable cut on clusters to be added to the event #1576 9c44139

Based on the summary of changes and comments above.

[cmsbuild]

This pull request is fully signed and it will be integrated in one of the next IBs unless changes or unless it breaks tests. @ktf can you please take care of it?